boegel / boegelbot

A bot that helps out with incoming contributions to GitHub repositories of the EasyBuild project

Bio

I'm Kenneth Hoste, a.k.a. 'boegel', a computer scientist and FOSS enthusiast from Belgium. I have a Masters (2005) and PhD (2010) in Computer Science from Ghent University. My dissertation topic was "Analysis, Estimation and Optimization of Computer System Performance Using Machine Learning".

Since October 2010, I'm a member of the HPC team at Ghent University (Belgium) where I am mainly responsible for user support & training. As a part of my job, I am also the lead developer and release manager of EasyBuild, a software build and installation framework for (scientific) software on High Performance Computing (HPC) systems.

In my free time, I am a family guy and a fan of loud music, frequently attending gigs and festivals.

I enjoy helping people & sharing his expertise, and like joking around. I have a weak spot for stickers and beer.

Things I don't like include CMake, Bazel, TensorFlow, C++, ...

• Twitter: @kehoste
• email: [email protected]

Projects

Current

• EasyBuild: A software build and installation framework for (scientific) software on HPC systems
  • lead developer & release manager (BDFL, if you will)
• archspec: A library for detecting, labeling, and reasoning about microarchitectures
  • co-developer
• EESSI: European Environment for Scientific Software Installations
  • active project partner
• PyGHee (pronounced as "piggy") is the GitHub Event Executor, a Python library to facilitate creating a GitHub App implemented in Python to process events from GitHub
  • creator and developer

Old

Projects I worked on during my PhD research:

• COLE: Compiler Optimization Level Exploration
  • GitHub repo: https://github.com/boegel/COLE
  • see CGO'08 paper: DOI, PDF, slides
• MICA:
  • GitHub repo: https://github.com/boegel/MICA
  • see IEEE Micro Hot Tutorials (2007): DOI, PDF

Talks

(some day I'll make this a complete list of talks I've given...)

2020

European Environment for Scientific Software Installations (EESSI) (HPCKP'20)

• date: June 18th 2020
• venue: HPC Knowledge Meeting - Barcelona (Spain) (virtual)
• slides
• recording (YouTube)

Easybuild State of the union (5th EasyBuild User Meeting)

• date: Jan 29th 2020
• venue: 5th EasyBuild User Meeting - Barcelona (Spain)
• slides
• recording (YouTube)

2019

Easybuild State of the union (4th EasyBuild User Meeting)

• date: Jan 30th 2019
• venue: 4th EasyBuild User Meeting - UCL, Louvain-la-Neuve (Belgium)
• slides
• recording (YouTube)

2018

In 2018, I gave 3 talks in a single weekend at FOSDEM'18 in Brussels, next to co-organizing a FOSDEM devroom and 3-day EasyBuild meetup the week beforre.

I highly recommend this experience to anyone (just kidding, please don't do this, it was nuts).

How To Make Package Managers Cry (FOSDEM'18)

This is probably the talk I'm most proud of so far (and I may never top it).

Born out of years of frustration to get scientific software installed...

• date: Feb 3rd 2018
• venue: FOSDEM'18 (Package Management devroom)
• FOSDEM page
• slides
• recording (YouTube)

Simplifying the contribution process for both contributors & maintainers (FOSDEM'18)

A case study of the GitHub integration in EasyBuild

• date: Feb 3rd 2018
• venue: FOSDEM'18 (Testing & Automation devroom)
• FOSDEM page
• slides
• recording (YouTube)

Installing software for scientists on a multi-user HPC system (FOSDEM'18)

A comparison between conda, EasyBuild, Guix, Nix & Spack

• date: Feb 3rd 2018
• venue: FOSDEM'18 (HPC, Big Data, and Data Science devroom)
• FOSDEM page
• slides
• recording (YouTube)

EasyBuild: past, present and future (3rd EasyBuild User Meeting)

• date: Jan 30th 2018
• venue: 3rd EasyBuild User Meeting - SURFsara, Amsterdam (Netherlands)
• slides
• recording (YouTube)

2017

EasyBuild: past, present and future (2nd EasyBuild User Meeting)

• date: Feb 8th 2017
• venue: 2nd EasyBuild User Meeting - Forschungszentrum Jülich (Germany)
• slides
• recording (YouTube)

2016

EasyBuild: past, present and future (1st EasyBuild User Meeting)

• date: Jan 27th 2016
• venue: 1st EasyBuild User Meeting - Ghent University (Belgium)
• slides

Why I like Lmod, and you should too (1st EasyBuild User Meeting)

• date: Jan 28th 2016
• venue: 1st EasyBuild User Meeting - Ghent University (Belgium)
• slides
• recording (YouTube)

Events

Here are some of the events I organized together with others:

2020

• 5th EasyBuild User Meeting (2020 - Barcelona, Spain)
• HPC, Big Data & Data Science devroom at FOSDEM'20 (2020 - Brussels, Belgium)

2019

• 4th EasyBuild User Meeting (2019 - Louvain-la-Neuve, Belgium)
• HPC, Big Data & Data Science devroom at FOSDEM'19 (2019 - Brussels, Belgium)

2018

• 3rd EasyBuild User Meeting (2018 - Amsterdam, Netherlands)
• HPC, Big Data & Data Science devroom at FOSDEM'18 (2018 - Brussels, Belgium)

2017

• 2nd EasyBuild User Meeting (2017 - Jülich, Germany)
• HPC, Big Data & Data Science devroom at FOSDEM'17 (2017 - Brussels, Belgium)

2016

• 1st EasyBuild User Meeting (2016 - Ghent, Belgium)
• HPC, Big Data & Data Science devroom at FOSDEM'16 (2016 - Brussels, Belgium)

2014

• HPC and computational science devroom at FOSDEM'14 (2014 - Brussels, Belgium)

boegelbot's Issues

Boegelbot should ask PR author to upload a test report

If a PR has not test report after a given time (e.g. 24h) boegelbot should kindly ask the PR author to upload a test report. In the message a link to the documentation of --upload-test-report could be added: https://docs.easybuild.io/en/latest/Integration_with_GitHub.html#updating-existing-pull-requests-update-pr

Teach Boegelbot about Containers

We might want to teach boegelbot how to run test reports using a Container.

automatically detect shebangs with hardcoded paths

this would catch errors more easily such as the one here easybuilders/easybuild-easyconfigs#17249 (comment)

quoting boegel:

but that should probably be opt-in, since we may hit false positives too, and the scanning may take a bit of time

boegelbot picking up quoted replies

As seen in easybuilders/easybuild-easyconfigs#18741 (comment), boegelbot will also trigger when @boegelbot is quoted. Perhaps we can check if the @ is inside a quote block, and ignore those?

Bot should present --review-pr in EasyConfig PR

If there are no labels

• If there is a brand spanking new easyconfig, add a new-label
• If there is an updated easyconfig versiom, add a update-label

Use git pull; eb --review-pr XXX --boegelbot-format-thing and format it in some nice way (which will need to be a bit different from the normal so it renders nicely in a diff codeblock

• Comparing to YYY-foss-2019a.eb

Line 10
- something
+ like this

Line 10:13
- checksums: [
-    'asdfasdfasdf
- ]
+ checksums = [
+    'asdfasdfasdf'
+    'asdfasdfasdf'
+]

I have no strong feeling about the format itself.

The bot should learn Travis CI allow_failures mode

@boegelbot should not complain about Travis CI jobs failing that are running in allow_failures mode.

hard crash with faulty user input via PR comment

[6/41] Processing notification for easybuild-easyconfigs PR #16913 "Add HiCBricks to Bioconductor 3.15"... (thread id: 5098448515, timestamp: 2022-12-15T20:59:36Z)
Found comment including '\s*@boegelbot:?\s*': @boegelbot please test @ generoso
CORE_CNT="--skip"
Comment includes '@.*generoso', so processing it...
Traceback (most recent call last):
  File "./boegelbot.py", line 688, in <module>
    main()
  File "./boegelbot.py", line 682, in main
    core_cnt)
  File "./boegelbot.py", line 570, in process_notifications
    (out, ec) = run_cmd(cmd, simple=False)
  File "/home/boegelbot/easybuild/easybuild-framework/easybuild/tools/run.py", line 89, in cache_aware_func
    res = func(cmd, *args, **kwargs)
  File "/home/boegelbot/easybuild/easybuild-framework/easybuild/tools/run.py", line 249, in run_cmd
    regexp=regexp, stream_output=stream_output, trace=trace)
  File "/home/boegelbot/easybuild/easybuild-framework/easybuild/tools/run.py", line 352, in complete_cmd
    return parse_cmd_output(cmd, stdouterr, ec, simple, log_all, log_ok, regexp)
  File "/home/boegelbot/easybuild/easybuild-framework/easybuild/tools/run.py", line 641, in parse_cmd_output
    raise EasyBuildError('cmd "%s" exited with exit code %s and output:\n%s', cmd, ec, stdouterr)
easybuild.tools.build_log.EasyBuildError: 'cmd "EB_PR=16913 EB_ARGS= EB_CONTAINER= /opt/software/slurm/bin/sbatch --job-name test_PR_16913 --ntasks="--skip" ~/boegelbot/eb_from_pr_upload_generoso.sh" exited with exit code 1 and output:\nsbatch: error: Invalid numeric value "--skip" for --ntasks.\n'

Boegelbot should close it's own comments as outdated

They are distracting when they are already fixed, boegelbot should just auto-hide-as-outdated when it does a new error report.

test installations for PRs should be done in /tmp, missing easyconfigs should be installed when updating to latest 'develop'